Structure_threader: an improved method for automation and parallelization of programs structure, fastStructure and MavericK on multicore CPU systems

Pina-Martins, Francisco; Silva, Diogo N.; Fino, Joana; Paulo, Octávio S.

http://hdl.handle.net/10400.18/5464

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
Structure_threader.pdf		523.79 KB	Adobe PDF	Download

Send Feedback

Authors

Pina-Martins, Francisco

Silva, Diogo N.

Fino, Joana

Paulo, Octávio S.

Abstract(s)

Structure_threader is a program to parallelize multiple runs of genetic clustering software that does not make use of multithreading technology (structure, fastStructure and MavericK) on multicore computers. Our approach was benchmarked across multiple systems and displayed great speed improvements relative to the single-threaded implementation, scaling very close to linearly with the number of physical cores used. Structure_threader was compared to previous software written for the same task-ParallelStructure and StrAuto and was proven to be the faster (up to 25% faster) wrapper under all tested scenarios. Furthermore, Structure_threader can perform several automatic and convenient operations, assisting the user in assessing the most biologically likely value of 'K' via implementations such as the "Evanno," or "Thermodynamic Integration" tests and automatically draw the "meanQ" plots (static or interactive) for each value of K (or even combined plots). Structure_threader is written in python 3 and licensed under the GPLv3. It can be downloaded free of charge at https://github.com/StuntsPT/Structure_threader.